选择方向强化学习的神经网络模型

A DIRECTION-SELECTIVE IMPOSING LEARNING MODEL IN NEURAL NETWORK

提出了神经网络模型的一种选择方向强化学习规则,定义并导出了新模型与Hopfield模型两种不同的筛选曲线,由此表明新模型对相关图样的分辨力优于Hopfield模型。在微机上模拟了由100个神经元构成的网络,结果显示新模型具有重复记忆这一神经生理学特点。定义并分行了记忆强度因子,模拟结果表明记忆强度因子愈大的记忆态,联想性能愈好,学习周期愈短。

This paper presents a new direction-selective imposing learning model in a neural network. The screening curve is defined. The differences between curve of new model and that of Hopfield model show the stroger ability of new model to resolve patterns than that of Hopfield model. When the new model consisting of 100 neurons was simulated on microcomputer, the results display that new model owns such neurophysiological property as repetitive memory. The memory intensity factor is defined and discussed. The simulation results demonstrated that the stronger the memory intensity, the hetter the associative performance and the shorter the training cycle.